I have my own VFS layer, and recently started trying to get the default page size to actually match the real device erase block size depending on the actual device the database is running on.
What I've found is that the xSectorSize method in the VFS is *not* called unless I unset the device characteristic of SQLITE_IOCAP_POWERSAFE_OVERWRITE (which is ironic since the device is power safe overwrite, but only if its the erase block size). This ends up meaning that the page_size always returns SQLITE_DEFAULT_PAGE_SIZE regardless of sector size. Obviously I could get around this by running pragma page_size prior to making the database in the first place, but that seems unnecessary considering that I already have a VFS layer in place.
